Really, really appreciate the article "When I use the eyex option of margins, what is it actually computing and how does it relate to the coefficients of the loglinear model?" I'm able to reproduce the eyex values by other means, but not able to get standard error values. I need to be able to rationalize a Stata model with external computations, and this is a real sticking point. I've read the information on the delta method, but based on that I get a different standard error than margins, eyex gives me. Is there another article that explains this? Thanks!